error bars when breaking the y-axis
 

Jon - I think this question is directed towards you. I have implemented a
version of your broken y-axis scheme. My data looks like the following:

sample average avedev
negative control 0.36252361 0.029793573
sample 1 3.265795907 2.878954305
sample 2 0.149275395 0.035304333
sample 3 73.57998574 2.237144009
sample 4 0.346937718 0.02837823
sample 5 4.636386221 0.089703534
sample 6 709.7039626 16.33742397
sample 7 1.157628293 0.667009325
negative control 0.494505996 0.051731733
RNase A 1pg/ml 1.604749215 0.106664119

As you can see, most of the sample are in the 0-5 range, but nos. 3 & 6 are
way out. The question I have is how do I plot the avedev as error bars when
I have broken the y-axis? I need to adjust these values to reflect the
change I have made to the average series. I have the y-axis broken twice,
with the sample #3 bar broken once and the sample #6 bar broken twice. I
also did not use the method you use to calculate the midpoint of the bar for
placement of the break; rather, I eyeballed it.

This might ba a case where a log scale on the Y axis is better. You wouldn't need to
break the axis (twice!) and the error bars will show comparative deviations as a
proportion of the average value.

If you keep with the broken axis chart, you have three bands: 0 to 7, 70 to 76, and
690 to 730. Around 3/4 of the chart will have to be used for the uppermost band,
which has only 1 point, while most of the points are crammed into the bottom 1/8 of
the chart. This distorts the data worse than keeping the axis unbroken.

On a log scale, in fact, points 1 and especially 7 show up as having remarkably wide
avedev values, while all other points (except for point 2) have spreads smaller than
the data markers in the chart.

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______

Wazooli wrote:

Jon - I think this question is directed towards you. I have implemented a
version of your broken y-axis scheme. My data looks like the following:

sample average avedev
negative control 0.36252361 0.029793573
sample 1 3.265795907 2.878954305
sample 2 0.149275395 0.035304333
sample 3 73.57998574 2.237144009
sample 4 0.346937718 0.02837823
sample 5 4.636386221 0.089703534
sample 6 709.7039626 16.33742397
sample 7 1.157628293 0.667009325
negative control 0.494505996 0.051731733
RNase A 1pg/ml 1.604749215 0.106664119

As you can see, most of the sample are in the 0-5 range, but nos. 3 & 6 are
way out. The question I have is how do I plot the avedev as error bars when
I have broken the y-axis? I need to adjust these values to reflect the
change I have made to the average series. I have the y-axis broken twice,
with the sample #3 bar broken once and the sample #6 bar broken twice. I
also did not use the method you use to calculate the midpoint of the bar for
placement of the break; rather, I eyeballed it.

I see what you're saying about changing the y-scale to log, as long as the
x-axis crosses at 0.1. This exercise was more for illustration purposes than
actually trying to glean anything meaningful. i did figure out a way to
scale the error bars to reflect the scaling of the data series however.
Thanks for your help.

"Jon Peltier" wrote:

This might ba a case where a log scale on the Y axis is better. You wouldn't need to
break the axis (twice!) and the error bars will show comparative deviations as a
proportion of the average value.

If you keep with the broken axis chart, you have three bands: 0 to 7, 70 to 76, and
690 to 730. Around 3/4 of the chart will have to be used for the uppermost band,
which has only 1 point, while most of the points are crammed into the bottom 1/8 of
the chart. This distorts the data worse than keeping the axis unbroken.

On a log scale, in fact, points 1 and especially 7 show up as having remarkably wide
avedev values, while all other points (except for point 2) have spreads smaller than
the data markers in the chart.

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______

Wazooli wrote:

Jon - I think this question is directed towards you. I have implemented a
version of your broken y-axis scheme. My data looks like the following:

sample average avedev
negative control 0.36252361 0.029793573
sample 1 3.265795907 2.878954305
sample 2 0.149275395 0.035304333
sample 3 73.57998574 2.237144009
sample 4 0.346937718 0.02837823
sample 5 4.636386221 0.089703534
sample 6 709.7039626 16.33742397
sample 7 1.157628293 0.667009325
negative control 0.494505996 0.051731733
RNase A 1pg/ml 1.604749215 0.106664119

As you can see, most of the sample are in the 0-5 range, but nos. 3 & 6 are
way out. The question I have is how do I plot the avedev as error bars when
I have broken the y-axis? I need to adjust these values to reflect the
change I have made to the average series. I have the y-axis broken twice,
with the sample #3 bar broken once and the sample #6 bar broken twice. I
also did not use the method you use to calculate the midpoint of the bar for
placement of the break; rather, I eyeballed it.

P.S. - is there any way to adjust the error bars so the negative error bar is
the same size as th epositive error bar when using a log y-axis scale?
Thanks again.

"Jon Peltier" wrote:

This might ba a case where a log scale on the Y axis is better. You wouldn't need to
break the axis (twice!) and the error bars will show comparative deviations as a
proportion of the average value.

If you keep with the broken axis chart, you have three bands: 0 to 7, 70 to 76, and
690 to 730. Around 3/4 of the chart will have to be used for the uppermost band,
which has only 1 point, while most of the points are crammed into the bottom 1/8 of
the chart. This distorts the data worse than keeping the axis unbroken.

On a log scale, in fact, points 1 and especially 7 show up as having remarkably wide
avedev values, while all other points (except for point 2) have spreads smaller than
the data markers in the chart.

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Peltier Technical Services
Tutorials and Custom Solutions
http://PeltierTech.com/
_______

Wazooli wrote:

Jon - I think this question is directed towards you. I have implemented a
version of your broken y-axis scheme. My data looks like the following:

sample average avedev
negative control 0.36252361 0.029793573
sample 1 3.265795907 2.878954305
sample 2 0.149275395 0.035304333
sample 3 73.57998574 2.237144009
sample 4 0.346937718 0.02837823
sample 5 4.636386221 0.089703534
sample 6 709.7039626 16.33742397
sample 7 1.157628293 0.667009325
negative control 0.494505996 0.051731733
RNase A 1pg/ml 1.604749215 0.106664119

As you can see, most of the sample are in the 0-5 range, but nos. 3 & 6 are
way out. The question I have is how do I plot the avedev as error bars when
I have broken the y-axis? I need to adjust these values to reflect the
change I have made to the average series. I have the y-axis broken twice,
with the sample #3 bar broken once and the sample #6 bar broken twice. I
also did not use the method you use to calculate the midpoint of the bar for
placement of the break; rather, I eyeballed it.